What is Mean Time to Recovery (MTTR)?
Definition
MTTR depends on two things happening quickly: detecting that a problem exists, and having a clear path to resolve it once detected. A team with fast detection but a slow, manual rollback process will still have a high MTTR, and the reverse is also true; a team that can roll back instantly but takes hours to notice something's wrong won't score well either. For database changes specifically, MTTR is closely tied to how quickly drift or an unexpected state gets flagged and how readily available a tested rollback is for the change that caused the problem. Both of those depend on infrastructure that exists before the incident, not something built during it.
Why MTTR matters
Failures happen even in well-run systems; what separates resilient teams from fragile ones is largely how fast they recover once something goes wrong. MTTR captures exactly that, and it's often a more meaningful signal than failure rate alone, since a team with a slightly higher failure rate but very fast recovery can still deliver better overall reliability than a team that fails less often but takes hours to fix it when they do. Customers generally notice how long an incident lasts far more than whether one happened at all.
How MTTR is measured
MTTR is calculated as the average elapsed time between a failure being detected and service being restored, whether that's through a rollback, a hotfix, or another remediation. For database changes, detection often depends on drift detection or monitoring catching an unexpected state quickly, and resolution depends on having a rollback ready to run rather than one that has to be written from scratch under pressure once the incident is already underway. A slow first ten minutes of detection can dominate the entire recovery window.
How Liquibase helps
Liquibase's rollback capabilities, generated automatically for many Change Types and testable ahead of time, are what turn a rollback decision into a fast, known action rather than an improvised one. Drift detection and Structured Logging shorten the detection half of MTTR by surfacing an unexpected database state quickly, and Liquibase Secure's targeted rollbacks let a team resolve an incident through the same governed process used for a normal deployment, rather than an unreviewed emergency workaround. Practicing a rollback in a lower environment before it's ever needed in production is one of the more reliable ways to keep MTTR low, since the steps are already familiar rather than being worked out for the first time during an actual incident.
